This is not a definitive answer to your inquiry, but I did want to take the opportunity to highlight a new upcoming capability in DataMiner that we are working on, and that's the GQI (Generic Query Interface).

This is going to be made available both as a standard API as well as a 'data source' that you can leverage in the new Dashboards module in a very intuitive manner (those new Dashboards will also have reporting capabilities, i.e. dispatching dashboard snapshots as PDF files).

We'll work on providing more details on the GQI, but in short: it is a very flexible graph-style of API, that you can leverage to query DataMiner for specific datasets.  To give you a simple use case example already covered by the current version:

A. you can ask for example for all elements of a specific type (e.g. give me all my Linux Servers - and it would return a list of all Servers).

B. you can then say that for all those, you would like to have the CPU load and the Memory Utilization and the OS Version (and it will return a table, with one row for each server, and a column for the value of each of the metrics you asked for).

C. you could then say that you would like to filter the list for a specific version of the OS (and it will return a table, where it removed all the servers that fall outside your filter).

D. you could then say that you would like to keep only the TOP-5 based on the CPU Load metric (and again it would update that table for you - note: each time you rerun the query, it would update the table for you find the servers that match your criteria).

E. you could then say for example that you want to have the average of Memory Utilization of all those servers that remain (and it would return one figure, being the average Memory Utilization of the top-5 servers based on CPU load, only considering the ones with the OS Version that you specified).

Now, this is a simple use case.  Because there are a lot of other options already supported by the GQI (such as performing aggregations such as SUM, Min, Max, Average, Count, Distinct Counts, Median, Percentiles, etc.).   A really powerful capability is that you can perform multiple of those queries and combine them (i.e. each query would give you a table of data, and if you have two of those you can then merge them together).   And of course the powerful proposition is that all data in DataMiner can be leveraged (e.g. if you use IDP in DataMiner, you could easily go and start filtering your servers in the use case above based on their location, their room, the rack they reside in, etc. Or leverage information on active services running, bookings, tickets, alarm states, etc.).

One thing that you mention that is not yet covered by the GQI is that you refer to a percentile across the time trace of a metric for a specified period (i.e. GQI supports a percentile across a specific metric in a column, but not across the time trace of a single metric).  But that's a great use case, so we definitely will take that on board.

Just wanted to share this already with you, and we'll get back asap on the options that you would have today without the GQI to get your use case covered.